Brendan Kitts

Bio: Brendan Kitts is an academic researcher from Microsoft. The author has contributed to research in topics: Click fraud & Online advertising. The author has an hindex of 17, co-authored 44 publications receiving 1084 citations.

Optimal Bidding on Keyword Auctions

Abstract: Pay per click (PPC) auctions are used to sell positions in search engines. These auctions have gained increasing commercial importance, and many companies offer software to bid on these auctions. We present a trading agent for PPC auctions that is the first in our knowledge to use an explicit profit objective function. The agent creates a look-ahead plan of its desired bids, which allows it to exhibit intelligent behaviour including the ability to hold back money during expensive periods. We tested the agent in the latter part of 2003 in a live Overture auction. The agent generated four times the number of visits as human managers, in addition to reducing cost and variability.

Visualization and analysis of user clickpaths

Abstract: Methods and data processing system readable media have been created to graph user clickstream data over a network or at a network site to yield meaningful and visually esthetic information. In one set of embodiments, the method can comprise (i) performing a significance test on data from a network log and generating significance results. The method can also comprise (ii) determining which of network addresses and clicktrails between network addresses meet a traffic flow criterion. The data that meet a significance criterion, traffic criterion, or both can form (iii) graphable addresses and relationships. The method can further comprise (iv) generating statistics about the graphable addresses and relationships. The method can still further comprise (v) generating a graph based on the statistics about the graphable addresses and relationships, and (vi) changing any or all of the traffic flow, significance criterion, and statistics being computed, and regenerating the graph.

Cross-sell: a fast promotion-tunable customer-item recommendation method based on conditionally independent probabilities

Abstract: We develop a method for recommending products to customers with applications to both on-line and surface mail promotional offers. Our method differs from previous work in collaborative filtering [8] and imputation [18], in that we assume probabilities are conditionally independent. This assumption, which is also made in Naive Bayes [5], enables us to pre-compute probabilities and store them in main memory, enabling very fast performance on millions of customers. The algorithm supports a variety of tunable parameters so that the method can address different promotional objectives. We tested the algorithm at an on-line hardware retailer, with 17,400 customers divided randomly into control and experimental groups. In the experimental group, clickthrough increased by +40% (p<0.01), revenue by +38% (p<0.07), and units sold by +61% (p<0.01). By changing the algorithm’s parameter settings we found that these results could be improved even further. This work demonstrates the considerable potential of automated data mining for dramatically increasing the profitability of on and off-line retail promotions.

Click-fraud protector

Abstract: Determining the probability that a user or program fraudulently initiated a web-page request is described herein. A data-mining component is configured to determine attributes associated with the web-page request. A computation component is configured to calculate a probability that the web-page request was fraudulently initiated. To calculate this probability, the attributes and other parameters are fed into a statistical model. An auction component is configured to locate one or more advertisements to display on the web page based on the probability. The auction component may also be configured to restrict the advertisements for display based on advertiser-specified target criteria.

Method of predicting a customer's business potential and a data processing system readable medium including code for the method

Abstract: A method can be used to predict the purchasing potential of customers. In one embodiment, the prediction can be based in part on transactional data that is routinely collected by many businesses. An item preference model, a maximum spending model, a geographic model, and any combination of them can be used to make the prediction. The item preference model can be based on which items the customer prefers based on transactional data. The maximum spending model can use the daily maximum spending amount for a customer to determine potential. The geographic model may be based on distance or geographic indicate. Using any or all of the models, if the customer is spending below his or her predicted potential, he or she may be targeted for offers or other promotions.

Item-based top-N recommendation algorithms

Abstract: The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems---a personalized information filtering technology used to identify a set of items that will be of interest to a certain user. User-based collaborative filtering is the most successful technology for building recommender systems to date and is extensively used in many commercial recommender systems. Unfortunately, the computational complexity of these methods grows linearly with the number of customers, which in typical commercial applications can be several millions. To address these scalability concerns model-based recommendation techniques have been developed. These techniques analyze the user--item matrix to discover relations between the different items and use these relations to compute the list of recommendations.In this article, we present one such class of model-based recommendation algorithms that first determines the similarities between the various items and then uses them to identify the set of items to be recommended. The key steps in this class of algorithms are (i) the method used to compute the similarity between the items, and (ii) the method used to combine these similarities in order to compute the similarity between a basket of items and a candidate recommender item. Our experimental evaluation on eight real datasets shows that these item-based algorithms are up to two orders of magnitude faster than the traditional user-neighborhood based recommender systems and provide recommendations with comparable or better quality.

Collaborative Filtering Recommender Systems

Abstract: Recommender systems are an important part of the information and e-commerce ecosystem. They represent a powerful method for enabling users to filter through large information and product spaces. Nearly two decades of research on collaborative filtering have led to a varied set of algorithms and a rich collection of tools for evaluating their performance. Research in the field is moving in the direction of a richer understanding of how recommender technology may be embedded in specific domains. The differing personalities exhibited by different recommender algorithms show that recommendation is not a one-size-fits-all problem. Specific tasks, information needs, and item domains represent unique problems for recommenders, and design and evaluation of recommenders needs to be done based on the user tasks to be supported. Effective deployments must begin with careful analysis of prospective users and their goals. Based on this analysis, system designers have a host of options for the choice of algorithm and for its embedding in the surrounding user experience. This paper discusses a~wide variety of the choices available and their implications, aiming to provide both practicioners and researchers with an introduction to the important issues underlying recommenders and current best practices for addressing these issues.

An MDP-Based Recommender System

Abstract: Typical recommender systems adopt a static view of the recommendation process and treat it as a prediction problem. We argue that it is more appropriate to view the problem of generating recommendations as a sequential optimization problem and, consequently, that Markov decision processes (MDPs) provide a more appropriate model for recommender systems. MDPs introduce two benefits: they take into account the long-term effects of each recommendation and the expected value of each recommendation. To succeed in practice, an MDP-based recommender system must employ a strong initial model, must be solvable quickly, and should not consume too much memory. In this paper, we describe our particular MDP model, its initialization using a predictive model, the solution and update algorithm, and its actual performance on a commercial site. We also describe the particular predictive model we used which outperforms previous models. Our system is one of a small number of commercially deployed recommender systems. As far as we know, it is the first to report experimental analysis conducted on a real commercial site. These results validate the commercial value of recommender systems, and in particular, of our MDP-based approach.

Evaluation of Item-Based Top-N Recommendation Algorithms

Abstract: The explosive growth of the world-wide-web and the emergence of e-commerce has led to the development of recommender systems---a personalized information filtering technology used to identify a set of N items that will be of interest to a certain user. User-based Collaborative filtering is the most successful technology for building recommender systems to date, and is extensively used in many commercial recommender systems. Unfortunately, the computational complexity of these methods grows linearly with the number of customers that in typical commercial applications can grow to be several millions. To address these scalability concerns item-based recommendation techniques have been developed that analyze the user-item matrix to identify relations between the different items, and use these relations to compute the list of recommendations.In this paper we present one such class of item-based recommendation algorithms that first determine the similarities between the various items and then used them to identify the set of items to be recommended. The key steps in this class of algorithms are (i) the method used to compute the similarity between the items, and (ii) the method used to combine these similarities in order to compute the similarity between a basket of items and a candidate recommender item. Our experimental evaluation on five different datasets show that the proposed item-based algorithms are up to 28 times faster than the traditional user-neighborhood based recommender systems and provide recommendations whose quality is up to 27% better.